2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
36 #include "opt_ipsec.h"
37 #include "opt_mbuf_stress_test.h"
38 #include "opt_mpath.h"
39 #include "opt_route.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
47 #include <sys/malloc.h>
51 #include <sys/protosw.h>
52 #include <sys/rmlock.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/sysctl.h>
57 #include <sys/ucred.h>
60 #include <net/if_var.h>
61 #include <net/if_llatbl.h>
62 #include <net/netisr.h>
64 #include <net/route.h>
65 #include <net/flowtable.h>
67 #include <net/radix_mpath.h>
69 #include <net/rss_config.h>
72 #include <netinet/in.h>
73 #include <netinet/in_kdtrace.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_pcb.h>
77 #include <netinet/in_rss.h>
78 #include <netinet/in_var.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_options.h>
82 #include <netinet/sctp.h>
83 #include <netinet/sctp_crc32.h>
86 #include <netipsec/ipsec_support.h>
88 #include <machine/in_cksum.h>
90 #include <security/mac/mac_framework.h>
92 #ifdef MBUF_STRESS_TEST
93 static int mbuf_frag_size = 0;
94 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
95 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
98 static void ip_mloopback(struct ifnet *, const struct mbuf *, int);
101 extern int in_mcast_loop;
102 extern struct protosw inetsw[];
105 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
106 struct sockaddr_in *dst, int *fibnum, int *error)
108 struct m_tag *fwd_tag = NULL;
114 ip = mtod(m, struct ip *);
116 /* Run through list of hooks for output packets. */
117 odst.s_addr = ip->ip_dst.s_addr;
118 *error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, 0, inp);
120 if ((*error) != 0 || m == NULL)
121 return 1; /* Finished */
123 ip = mtod(m, struct ip *);
125 /* See if destination IP address was changed by packet filter. */
126 if (odst.s_addr != ip->ip_dst.s_addr) {
127 m->m_flags |= M_SKIP_FIREWALL;
128 /* If destination is now ourself drop to ip_input(). */
129 if (in_localip(ip->ip_dst)) {
130 m->m_flags |= M_FASTFWD_OURS;
131 if (m->m_pkthdr.rcvif == NULL)
132 m->m_pkthdr.rcvif = V_loif;
133 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
134 m->m_pkthdr.csum_flags |=
135 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
136 m->m_pkthdr.csum_data = 0xffff;
138 m->m_pkthdr.csum_flags |=
139 CSUM_IP_CHECKED | CSUM_IP_VALID;
141 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
142 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
144 *error = netisr_queue(NETISR_IP, m);
145 return 1; /* Finished */
148 bzero(dst, sizeof(*dst));
149 dst->sin_family = AF_INET;
150 dst->sin_len = sizeof(*dst);
151 dst->sin_addr = ip->ip_dst;
153 return -1; /* Reloop */
155 /* See if fib was changed by packet filter. */
156 if ((*fibnum) != M_GETFIB(m)) {
157 m->m_flags |= M_SKIP_FIREWALL;
158 *fibnum = M_GETFIB(m);
159 return -1; /* Reloop for FIB change */
162 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
163 if (m->m_flags & M_FASTFWD_OURS) {
164 if (m->m_pkthdr.rcvif == NULL)
165 m->m_pkthdr.rcvif = V_loif;
166 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
167 m->m_pkthdr.csum_flags |=
168 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
169 m->m_pkthdr.csum_data = 0xffff;
172 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
173 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
175 m->m_pkthdr.csum_flags |=
176 CSUM_IP_CHECKED | CSUM_IP_VALID;
178 *error = netisr_queue(NETISR_IP, m);
179 return 1; /* Finished */
181 /* Or forward to some other address? */
182 if ((m->m_flags & M_IP_NEXTHOP) &&
183 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
184 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
185 m->m_flags |= M_SKIP_FIREWALL;
186 m->m_flags &= ~M_IP_NEXTHOP;
187 m_tag_delete(m, fwd_tag);
189 return -1; /* Reloop for CHANGE of dst */
196 * IP output. The packet in mbuf chain m contains a skeletal IP
197 * header (with len, off, ttl, proto, tos, src, dst).
198 * The mbuf chain containing the packet will be freed.
199 * The mbuf opt, if present, will not be freed.
200 * If route ro is present and has ro_rt initialized, route lookup would be
201 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
202 * then result of route lookup is stored in ro->ro_rt.
204 * In the IP forwarding case, the packet will arrive with options already
205 * inserted, so must have a NULL opt pointer.
208 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
209 struct ip_moptions *imo, struct inpcb *inp)
211 struct rm_priotracker in_ifa_tracker;
213 struct ifnet *ifp = NULL; /* keep compiler happy */
215 int hlen = sizeof (struct ip);
218 struct sockaddr_in *dst;
219 const struct sockaddr_in *gw;
220 struct in_ifaddr *ia;
222 uint16_t ip_len, ip_off;
223 struct route iproute;
224 struct rtentry *rte; /* cache for ro->ro_rt */
227 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
228 int no_route_but_check_spd = 0;
233 INP_LOCK_ASSERT(inp);
234 M_SETFIB(m, inp->inp_inc.inc_fibnum);
235 if ((flags & IP_NODEFAULTFLOWID) == 0) {
236 m->m_pkthdr.flowid = inp->inp_flowid;
237 M_HASHTYPE_SET(m, inp->inp_flowtype);
243 bzero(ro, sizeof (*ro));
247 if (ro->ro_rt == NULL)
248 (void )flowtable_lookup(AF_INET, m, ro);
253 m = ip_insertoptions(m, opt, &len);
255 hlen = len; /* ip->ip_hl is updated above */
257 ip = mtod(m, struct ip *);
258 ip_len = ntohs(ip->ip_len);
259 ip_off = ntohs(ip->ip_off);
261 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
262 ip->ip_v = IPVERSION;
263 ip->ip_hl = hlen >> 2;
266 /* Header already set, fetch hlen from there */
267 hlen = ip->ip_hl << 2;
269 if ((flags & IP_FORWARDING) == 0)
270 IPSTAT_INC(ips_localout);
275 * dst can be rewritten but always points to &ro->ro_dst.
276 * gw is readonly but can point either to dst OR rt_gateway,
277 * therefore we need restore gw if we're redoing lookup.
279 gw = dst = (struct sockaddr_in *)&ro->ro_dst;
280 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
283 bzero(dst, sizeof(*dst));
284 dst->sin_family = AF_INET;
285 dst->sin_len = sizeof(*dst);
286 dst->sin_addr = ip->ip_dst;
290 * Validate route against routing table additions;
291 * a better/more specific route might have been added.
294 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
296 * If there is a cached route,
297 * check that it is to the same destination
298 * and is still up. If not, free it and try again.
299 * The address family should also be checked in case of sharing the
301 * Also check whether routing cache needs invalidation.
304 if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
305 rte->rt_ifp == NULL ||
306 !RT_LINK_IS_UP(rte->rt_ifp) ||
307 dst->sin_family != AF_INET ||
308 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
310 rte = ro->ro_rt = (struct rtentry *)NULL;
312 LLE_FREE(ro->ro_lle); /* zeros ro_lle */
313 ro->ro_lle = (struct llentry *)NULL;
318 * If routing to interface only, short circuit routing lookup.
319 * The use of an all-ones broadcast address implies this; an
320 * interface is specified by the broadcast address of an interface,
321 * or the destination address of a ptp interface.
323 if (flags & IP_SENDONES) {
324 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
325 M_GETFIB(m)))) == NULL &&
326 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
327 M_GETFIB(m)))) == NULL) {
328 IPSTAT_INC(ips_noroute);
333 ip->ip_dst.s_addr = INADDR_BROADCAST;
334 dst->sin_addr = ip->ip_dst;
338 } else if (flags & IP_ROUTETOIF) {
339 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
340 M_GETFIB(m)))) == NULL &&
341 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
342 M_GETFIB(m)))) == NULL) {
343 IPSTAT_INC(ips_noroute);
350 isbroadcast = in_broadcast(dst->sin_addr, ifp);
351 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
352 imo != NULL && imo->imo_multicast_ifp != NULL) {
354 * Bypass the normal routing lookup for multicast
355 * packets if the interface is specified.
357 ifp = imo->imo_multicast_ifp;
358 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
361 isbroadcast = 0; /* fool gcc */
364 * We want to do any cloning requested by the link layer,
365 * as this is probably required in all cases for correct
366 * operation (as it is for ARP).
370 rtalloc_mpath_fib(ro,
371 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
374 in_rtalloc_ign(ro, 0, fibnum);
379 (rte->rt_flags & RTF_UP) == 0 ||
380 rte->rt_ifp == NULL ||
381 !RT_LINK_IS_UP(rte->rt_ifp)) {
382 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
384 * There is no route for this packet, but it is
385 * possible that a matching SPD entry exists.
387 no_route_but_check_spd = 1;
388 mtu = 0; /* Silence GCC warning. */
391 IPSTAT_INC(ips_noroute);
392 error = EHOSTUNREACH;
395 ia = ifatoia(rte->rt_ifa);
397 counter_u64_add(rte->rt_pksent, 1);
398 rt_update_ro_flags(ro);
399 if (rte->rt_flags & RTF_GATEWAY)
400 gw = (struct sockaddr_in *)rte->rt_gateway;
401 if (rte->rt_flags & RTF_HOST)
402 isbroadcast = (rte->rt_flags & RTF_BROADCAST);
404 isbroadcast = in_broadcast(gw->sin_addr, ifp);
408 * Calculate MTU. If we have a route that is up, use that,
409 * otherwise use the interface's MTU.
411 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
415 /* Catch a possible divide by zero later. */
416 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
417 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
419 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
420 m->m_flags |= M_MCAST;
422 * IP destination address is multicast. Make sure "gw"
423 * still points to the address in "ro". (It may have been
424 * changed to point to a gateway address, above.)
428 * See if the caller provided any multicast options
431 ip->ip_ttl = imo->imo_multicast_ttl;
432 if (imo->imo_multicast_vif != -1)
435 ip_mcast_src(imo->imo_multicast_vif) :
438 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
440 * Confirm that the outgoing interface supports multicast.
442 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
443 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
444 IPSTAT_INC(ips_noroute);
450 * If source address not specified yet, use address
451 * of outgoing interface.
453 if (ip->ip_src.s_addr == INADDR_ANY) {
454 /* Interface may have no addresses. */
456 ip->ip_src = IA_SIN(ia)->sin_addr;
459 if ((imo == NULL && in_mcast_loop) ||
460 (imo && imo->imo_multicast_loop)) {
462 * Loop back multicast datagram if not expressly
463 * forbidden to do so, even if we are not a member
464 * of the group; ip_input() will filter it later,
465 * thus deferring a hash lookup and mutex acquisition
466 * at the expense of a cheap copy using m_copym().
468 ip_mloopback(ifp, m, hlen);
471 * If we are acting as a multicast router, perform
472 * multicast forwarding as if the packet had just
473 * arrived on the interface to which we are about
474 * to send. The multicast forwarding function
475 * recursively calls this function, using the
476 * IP_FORWARDING flag to prevent infinite recursion.
478 * Multicasts that are looped back by ip_mloopback(),
479 * above, will be forwarded by the ip_input() routine,
482 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
484 * If rsvp daemon is not running, do not
485 * set ip_moptions. This ensures that the packet
486 * is multicast and not just sent down one link
487 * as prescribed by rsvpd.
492 ip_mforward(ip, ifp, m, imo) != 0) {
500 * Multicasts with a time-to-live of zero may be looped-
501 * back, above, but must not be transmitted on a network.
502 * Also, multicasts addressed to the loopback interface
503 * are not sent -- the above call to ip_mloopback() will
504 * loop back a copy. ip_input() will drop the copy if
505 * this host does not belong to the destination group on
506 * the loopback interface.
508 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
517 * If the source address is not specified yet, use the address
518 * of the outoing interface.
520 if (ip->ip_src.s_addr == INADDR_ANY) {
521 /* Interface may have no addresses. */
523 ip->ip_src = IA_SIN(ia)->sin_addr;
528 * Look for broadcast address and
529 * verify user is allowed to send
533 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
534 error = EADDRNOTAVAIL;
537 if ((flags & IP_ALLOWBROADCAST) == 0) {
541 /* don't allow broadcast messages to be fragmented */
546 m->m_flags |= M_BCAST;
548 m->m_flags &= ~M_BCAST;
552 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
553 if (IPSEC_ENABLED(ipv4)) {
554 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
555 if (error == EINPROGRESS)
561 * Check if there was a route for this packet; return error if not.
563 if (no_route_but_check_spd) {
564 IPSTAT_INC(ips_noroute);
565 error = EHOSTUNREACH;
568 /* Update variables that are affected by ipsec4_output(). */
569 ip = mtod(m, struct ip *);
570 hlen = ip->ip_hl << 2;
573 /* Jump over all PFIL processing if hooks are not active. */
574 if (PFIL_HOOKED(&V_inet_pfil_hook)) {
575 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
576 case 1: /* Finished */
579 case 0: /* Continue normally */
580 ip = mtod(m, struct ip *);
583 case -1: /* Need to try again */
584 /* Reset everything for a new round */
587 ifa_free(&ia->ia_ifa);
588 ro->ro_prepend = NULL;
591 ip = mtod(m, struct ip *);
597 /* 127/8 must not appear on wire - RFC1122. */
598 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
599 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
600 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
601 IPSTAT_INC(ips_badaddr);
602 error = EADDRNOTAVAIL;
607 m->m_pkthdr.csum_flags |= CSUM_IP;
608 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
610 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
613 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
614 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
615 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
620 * If small enough for interface, or the interface will take
621 * care of the fragmentation for us, we can just send directly.
624 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
626 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
627 ip->ip_sum = in_cksum(m, hlen);
628 m->m_pkthdr.csum_flags &= ~CSUM_IP;
632 * Record statistics for this interface address.
633 * With CSUM_TSO the byte/packet count will be slightly
634 * incorrect because we count the IP+TCP headers only
635 * once instead of for every generated packet.
637 if (!(flags & IP_FORWARDING) && ia) {
638 if (m->m_pkthdr.csum_flags & CSUM_TSO)
639 counter_u64_add(ia->ia_ifa.ifa_opackets,
640 m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
642 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
644 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
646 #ifdef MBUF_STRESS_TEST
647 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
648 m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
651 * Reset layer specific mbuf flags
652 * to avoid confusing lower layers.
655 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
656 error = (*ifp->if_output)(ifp, m,
657 (const struct sockaddr *)gw, ro);
661 /* Balk when DF bit is set or the interface didn't support TSO. */
662 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
664 IPSTAT_INC(ips_cantfrag);
669 * Too large for interface; fragment if possible. If successful,
670 * on return, m will point to a list of packets to be sent.
672 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
679 /* Record statistics for this interface address. */
681 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
682 counter_u64_add(ia->ia_ifa.ifa_obytes,
686 * Reset layer specific mbuf flags
687 * to avoid confusing upper layers.
691 IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
692 mtod(m, struct ip *), NULL);
693 error = (*ifp->if_output)(ifp, m,
694 (const struct sockaddr *)gw, ro);
700 IPSTAT_INC(ips_fragmented);
704 * Release the route if using our private route, or if
705 * (with flowtable) we don't have our own reference.
707 if (ro == &iproute || ro->ro_flags & RT_NORTREF)
709 else if (rte == NULL)
711 * If the caller supplied a route but somehow the reference
712 * to it has been released need to prevent the caller
713 * calling RTFREE on it again.
717 ifa_free(&ia->ia_ifa);
725 * Create a chain of fragments which fit the given mtu. m_frag points to the
726 * mbuf to be fragmented; on return it points to the chain with the fragments.
727 * Return 0 if no error. If error, m_frag may contain a partially built
728 * chain of fragments that should be freed by the caller.
730 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
733 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
734 u_long if_hwassist_flags)
737 int hlen = ip->ip_hl << 2;
738 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
740 struct mbuf *m0 = *m_frag; /* the original packet */
744 uint16_t ip_len, ip_off;
746 ip_len = ntohs(ip->ip_len);
747 ip_off = ntohs(ip->ip_off);
749 if (ip_off & IP_DF) { /* Fragmentation not allowed */
750 IPSTAT_INC(ips_cantfrag);
755 * Must be able to put at least 8 bytes per fragment.
761 * If the interface will not calculate checksums on
762 * fragmented packets, then do it here.
764 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
765 in_delayed_cksum(m0);
766 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
769 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
770 sctp_delayed_cksum(m0, hlen);
771 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
774 if (len > PAGE_SIZE) {
776 * Fragment large datagrams such that each segment
777 * contains a multiple of PAGE_SIZE amount of data,
778 * plus headers. This enables a receiver to perform
779 * page-flipping zero-copy optimizations.
781 * XXX When does this help given that sender and receiver
782 * could have different page sizes, and also mtu could
783 * be less than the receiver's page size ?
787 off = MIN(mtu, m0->m_pkthdr.len);
790 * firstlen (off - hlen) must be aligned on an
794 goto smart_frag_failure;
795 off = ((off - hlen) & ~7) + hlen;
796 newlen = (~PAGE_MASK) & mtu;
797 if ((newlen + sizeof (struct ip)) > mtu) {
798 /* we failed, go back the default */
809 firstlen = off - hlen;
810 mnext = &m0->m_nextpkt; /* pointer to next packet */
813 * Loop through length of segment after first fragment,
814 * make new header and copy data of each part and link onto chain.
815 * Here, m0 is the original packet, m is the fragment being created.
816 * The fragments are linked off the m_nextpkt of the original
817 * packet, which after processing serves as the first fragment.
819 for (nfrags = 1; off < ip_len; off += len, nfrags++) {
820 struct ip *mhip; /* ip header on the fragment */
822 int mhlen = sizeof (struct ip);
824 m = m_gethdr(M_NOWAIT, MT_DATA);
827 IPSTAT_INC(ips_odropped);
831 * Make sure the complete packet header gets copied
832 * from the originating mbuf to the newly created
833 * mbuf. This also ensures that existing firewall
834 * classification(s), VLAN tags and so on get copied
835 * to the resulting fragmented packet(s):
837 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
840 IPSTAT_INC(ips_odropped);
844 * In the first mbuf, leave room for the link header, then
845 * copy the original IP header including options. The payload
846 * goes into an additional mbuf chain returned by m_copym().
848 m->m_data += max_linkhdr;
849 mhip = mtod(m, struct ip *);
851 if (hlen > sizeof (struct ip)) {
852 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
853 mhip->ip_v = IPVERSION;
854 mhip->ip_hl = mhlen >> 2;
857 /* XXX do we need to add ip_off below ? */
858 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
859 if (off + len >= ip_len)
862 mhip->ip_off |= IP_MF;
863 mhip->ip_len = htons((u_short)(len + mhlen));
864 m->m_next = m_copym(m0, off, len, M_NOWAIT);
865 if (m->m_next == NULL) { /* copy failed */
867 error = ENOBUFS; /* ??? */
868 IPSTAT_INC(ips_odropped);
871 m->m_pkthdr.len = mhlen + len;
873 mac_netinet_fragment(m0, m);
875 mhip->ip_off = htons(mhip->ip_off);
877 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
878 mhip->ip_sum = in_cksum(m, mhlen);
879 m->m_pkthdr.csum_flags &= ~CSUM_IP;
882 mnext = &m->m_nextpkt;
884 IPSTAT_ADD(ips_ofragments, nfrags);
887 * Update first fragment by trimming what's been copied out
888 * and updating header.
890 m_adj(m0, hlen + firstlen - ip_len);
891 m0->m_pkthdr.len = hlen + firstlen;
892 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
893 ip->ip_off = htons(ip_off | IP_MF);
895 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
896 ip->ip_sum = in_cksum(m0, hlen);
897 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
906 in_delayed_cksum(struct mbuf *m)
909 uint16_t csum, offset, ip_len;
911 ip = mtod(m, struct ip *);
912 offset = ip->ip_hl << 2 ;
913 ip_len = ntohs(ip->ip_len);
914 csum = in_cksum_skip(m, ip_len, offset);
915 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
917 offset += m->m_pkthdr.csum_data; /* checksum offset */
919 /* find the mbuf in the chain where the checksum starts*/
920 while ((m != NULL) && (offset >= m->m_len)) {
924 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
925 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
926 *(u_short *)(m->m_data + offset) = csum;
930 * IP socket option processing.
933 ip_ctloutput(struct socket *so, struct sockopt *sopt)
935 struct inpcb *inp = sotoinpcb(so);
943 if (sopt->sopt_level != IPPROTO_IP) {
946 if (sopt->sopt_level == SOL_SOCKET &&
947 sopt->sopt_dir == SOPT_SET) {
948 switch (sopt->sopt_name) {
951 if ((so->so_options & SO_REUSEADDR) != 0)
952 inp->inp_flags2 |= INP_REUSEADDR;
954 inp->inp_flags2 &= ~INP_REUSEADDR;
960 if ((so->so_options & SO_REUSEPORT) != 0)
961 inp->inp_flags2 |= INP_REUSEPORT;
963 inp->inp_flags2 &= ~INP_REUSEPORT;
969 inp->inp_inc.inc_fibnum = so->so_fibnum;
980 switch (sopt->sopt_dir) {
982 switch (sopt->sopt_name) {
989 if (sopt->sopt_valsize > MLEN) {
993 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
998 m->m_len = sopt->sopt_valsize;
999 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1006 error = ip_pcbopts(inp, sopt->sopt_name, m);
1012 if (sopt->sopt_td != NULL) {
1013 error = priv_check(sopt->sopt_td,
1014 PRIV_NETINET_BINDANY);
1021 case IP_RSS_LISTEN_BUCKET:
1027 case IP_RECVRETOPTS:
1028 case IP_RECVDSTADDR:
1036 case IP_RECVRSSBUCKETID:
1038 error = sooptcopyin(sopt, &optval, sizeof optval,
1043 switch (sopt->sopt_name) {
1045 inp->inp_ip_tos = optval;
1049 inp->inp_ip_ttl = optval;
1053 if (optval >= 0 && optval <= MAXTTL)
1054 inp->inp_ip_minttl = optval;
1059 #define OPTSET(bit) do { \
1062 inp->inp_flags |= bit; \
1064 inp->inp_flags &= ~bit; \
1068 #define OPTSET2(bit, val) do { \
1071 inp->inp_flags2 |= bit; \
1073 inp->inp_flags2 &= ~bit; \
1078 OPTSET(INP_RECVOPTS);
1081 case IP_RECVRETOPTS:
1082 OPTSET(INP_RECVRETOPTS);
1085 case IP_RECVDSTADDR:
1086 OPTSET(INP_RECVDSTADDR);
1090 OPTSET(INP_RECVTTL);
1098 OPTSET(INP_ONESBCAST);
1101 OPTSET(INP_DONTFRAG);
1104 OPTSET(INP_BINDANY);
1107 OPTSET(INP_RECVTOS);
1110 OPTSET2(INP_BINDMULTI, optval);
1113 OPTSET2(INP_RECVFLOWID, optval);
1116 case IP_RSS_LISTEN_BUCKET:
1117 if ((optval >= 0) &&
1118 (optval < rss_getnumbuckets())) {
1119 inp->inp_rss_listen_bucket = optval;
1120 OPTSET2(INP_RSS_BUCKET_SET, 1);
1125 case IP_RECVRSSBUCKETID:
1126 OPTSET2(INP_RECVRSSBUCKETID, optval);
1135 * Multicast socket options are processed by the in_mcast
1138 case IP_MULTICAST_IF:
1139 case IP_MULTICAST_VIF:
1140 case IP_MULTICAST_TTL:
1141 case IP_MULTICAST_LOOP:
1142 case IP_ADD_MEMBERSHIP:
1143 case IP_DROP_MEMBERSHIP:
1144 case IP_ADD_SOURCE_MEMBERSHIP:
1145 case IP_DROP_SOURCE_MEMBERSHIP:
1146 case IP_BLOCK_SOURCE:
1147 case IP_UNBLOCK_SOURCE:
1149 case MCAST_JOIN_GROUP:
1150 case MCAST_LEAVE_GROUP:
1151 case MCAST_JOIN_SOURCE_GROUP:
1152 case MCAST_LEAVE_SOURCE_GROUP:
1153 case MCAST_BLOCK_SOURCE:
1154 case MCAST_UNBLOCK_SOURCE:
1155 error = inp_setmoptions(inp, sopt);
1159 error = sooptcopyin(sopt, &optval, sizeof optval,
1166 case IP_PORTRANGE_DEFAULT:
1167 inp->inp_flags &= ~(INP_LOWPORT);
1168 inp->inp_flags &= ~(INP_HIGHPORT);
1171 case IP_PORTRANGE_HIGH:
1172 inp->inp_flags &= ~(INP_LOWPORT);
1173 inp->inp_flags |= INP_HIGHPORT;
1176 case IP_PORTRANGE_LOW:
1177 inp->inp_flags &= ~(INP_HIGHPORT);
1178 inp->inp_flags |= INP_LOWPORT;
1188 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1189 case IP_IPSEC_POLICY:
1190 if (IPSEC_ENABLED(ipv4)) {
1191 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1198 error = ENOPROTOOPT;
1204 switch (sopt->sopt_name) {
1207 if (inp->inp_options)
1208 error = sooptcopyout(sopt,
1209 mtod(inp->inp_options,
1211 inp->inp_options->m_len);
1213 sopt->sopt_valsize = 0;
1220 case IP_RECVRETOPTS:
1221 case IP_RECVDSTADDR:
1234 case IP_RSSBUCKETID:
1235 case IP_RECVRSSBUCKETID:
1237 switch (sopt->sopt_name) {
1240 optval = inp->inp_ip_tos;
1244 optval = inp->inp_ip_ttl;
1248 optval = inp->inp_ip_minttl;
1251 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1252 #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
1255 optval = OPTBIT(INP_RECVOPTS);
1258 case IP_RECVRETOPTS:
1259 optval = OPTBIT(INP_RECVRETOPTS);
1262 case IP_RECVDSTADDR:
1263 optval = OPTBIT(INP_RECVDSTADDR);
1267 optval = OPTBIT(INP_RECVTTL);
1271 optval = OPTBIT(INP_RECVIF);
1275 if (inp->inp_flags & INP_HIGHPORT)
1276 optval = IP_PORTRANGE_HIGH;
1277 else if (inp->inp_flags & INP_LOWPORT)
1278 optval = IP_PORTRANGE_LOW;
1284 optval = OPTBIT(INP_ONESBCAST);
1287 optval = OPTBIT(INP_DONTFRAG);
1290 optval = OPTBIT(INP_BINDANY);
1293 optval = OPTBIT(INP_RECVTOS);
1296 optval = inp->inp_flowid;
1299 optval = inp->inp_flowtype;
1302 optval = OPTBIT2(INP_RECVFLOWID);
1305 case IP_RSSBUCKETID:
1306 retval = rss_hash2bucket(inp->inp_flowid,
1310 optval = rss_bucket;
1314 case IP_RECVRSSBUCKETID:
1315 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1319 optval = OPTBIT2(INP_BINDMULTI);
1322 error = sooptcopyout(sopt, &optval, sizeof optval);
1326 * Multicast socket options are processed by the in_mcast
1329 case IP_MULTICAST_IF:
1330 case IP_MULTICAST_VIF:
1331 case IP_MULTICAST_TTL:
1332 case IP_MULTICAST_LOOP:
1334 error = inp_getmoptions(inp, sopt);
1337 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1338 case IP_IPSEC_POLICY:
1339 if (IPSEC_ENABLED(ipv4)) {
1340 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1347 error = ENOPROTOOPT;
1356 * Routine called from ip_output() to loop back a copy of an IP multicast
1357 * packet to the input queue of a specified interface. Note that this
1358 * calls the output routine of the loopback "driver", but with an interface
1359 * pointer that might NOT be a loopback interface -- evil, but easier than
1360 * replicating that code here.
1363 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1369 * Make a deep copy of the packet because we're going to
1370 * modify the pack in order to generate checksums.
1372 copym = m_dup(m, M_NOWAIT);
1373 if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1374 copym = m_pullup(copym, hlen);
1375 if (copym != NULL) {
1376 /* If needed, compute the checksum and mark it as valid. */
1377 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1378 in_delayed_cksum(copym);
1379 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1380 copym->m_pkthdr.csum_flags |=
1381 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1382 copym->m_pkthdr.csum_data = 0xffff;
1385 * We don't bother to fragment if the IP length is greater
1386 * than the interface's MTU. Can this possibly matter?
1388 ip = mtod(copym, struct ip *);
1390 ip->ip_sum = in_cksum(copym, hlen);
1391 if_simloop(ifp, copym, AF_INET, 0);